Migrating from V1Beta2

Document AI's v1beta2 API is now deprecated in favor of the v1 API.

Unlike v1beta2, Document AI v1 is a stateful API, meaning you need to create a Processor before you can send requests to Process and BatchProcess documents.

NOTE: This deprecation only affects the version of the API used to call any processor. It does not affect the availability of any processor.

The major areas of v1 API backward incompatibility with v1beta2 API are discussed later in this document.

Creating a processor

To migrate to v1 for processor creation, first create a Processor. The following table serves as a general guideline for choosing an appropriate ProcessorType to create, but you can consult the full processor list.

document_type params Processor type
invoice INVOICE_PROCESSOR
general OCR_PROCESSOR
general table_extraction_params FORM_PARSER_PROCESSOR
general form_extraction_params FORM_PARSER_PROCESSOR
general entity_extraction_params FORM_PARSER_PROCESSOR
general automl_params CUSTOM_EXTRACTION_PROCESSOR

For a step-by-step guide to creating a processor, view the guide.

Processing documents

Document AI v1 has two APIs to accept documents: Process and BatchProcess. Each of these methods corresponds to a method from v1beta2. The methods and their differences are highlighted:

v1beta2 Method v1 Method Migration notes
ProcessDocument Process InputConfig.Cloud Storage_source not supported in v1, use BatchProcess to process documents on Cloud Storage.
BatchProcessDocuments BatchProcess In addition to accepting individual documents on Cloud Storage, v1 supports a Cloud Storage prefix to process many documents at once.

Client libraries

If using the Client Libraries to interact with Document AI, switch from using the DocumentUnderstandingService to the DocumentProcessorService. For a detailed example of using the client libraries for v1, see client libraries.